Evaluating and Comparing Systemic Risk and Market Risk of

Evaluating and Comparing Systemic Risk and Market Risk of

Mutual Funds in Iran Capital Market

Fereshteh Shahbazin Ph.D. Candidate of Department of Financial Management, Qazvin Branch, Islamic

Azad University, Qazvin, Iran (Email: [email protected])

Hasan Ghalibaf Asl*

*Corresponding author, Prof., Department of Finance and Insurance, Faculty of

Management, Alzahra University, Tehran, Iran. (Email: [email protected])

Mohsen Seighali Assistant Prof., Faculty of Accounting and Management, Islamic Azad University,

Qazvin, Iran. (Email: [email protected])

Moslem Peymani Foroushani

Assistant Prof., Faculty of Finance and Banking, Allameh Tabatabaei University,

Tehran, Iran. (Email: [email protected])

Document Type: Original Article 2019, Vol. 3, No. 4. 90-112

Received: 2020/05/15 Accepted: 2020/07/10 Published: 2020/08/10

Abstract Mutual funds are one of the most paramount investment mechanisms in financial markets. By playing a financial intermediary role, they give nonprofessionals access to professionally managed portfolios of securities and provide numerous benefits for both the capital market and investors simultaneously. This study evaluated and investigated the systemic risk of mutual funds in the Iran capital market by adopting a Conditional Value at Risk (CoVaR) approach and employing quantile regression. In the finance literature, systemic risk is the probability of a downfall in the financial system when a segment or an individual component gets in distress. This risk can trigger instability or shock in financial markets and the real part of the economy. The results revealed that stock (equity) mutual funds were systemically more important than other funds, including fixed-income and balanced mutual funds, due to the high volatility in their return, which makes them riskier. To compare systemic risk and market risk among mutual funds, funds classified into five different groups based on their systemic risk. According to this categorization, analysis of variance illuminated that the market risk of mutual funds had a direct relationship with their systemic risk,

such that a higher systemic risk of a fund stood for higher market risk. Keywords: Conditional Value at Risk, Mutual Funds, Quantile Regression, Systemic

Risk

DOI: 10.22034/ijf.2020.231342.1127 © Iran Finance Association

http://ifsa.org.ir/

http://ifsa.org.ir/

91

Evaluating and Comparing Systemic Risk and Market Risk

Introduction

With the enlargement of the capital market, its operational and regulatory

structure should inevitably change, and the coordination among its components

has to increase. In the development process of the capital market, the mere

enlargement of its components does not suffice, since it decreases the

efficiency of the entire system in the long run. Constant development and

growth of a system entail the gradual modification, extension, or empowerment

of system components, including institutions, mechanisms, and processes. A

reflection on the recent financial crisis doctrine highlights the need for the

empowerment and modification of regulatory structures, along with attention to

the systemic risk in financial markets.

Systemic risk in finance literature means the likelihood of a downfall in

the whole financial system. This risk can trigger instability or distress in

financial markets. One of the important issues in the systemic risk is a

contagion, which is the spreading probability of significant economic changes

in a sector or an institution to another. The banking crises of the previous

decades, and the 2007-2009 financial crises at their top, made the systemic risk

debate in the financial markets to be noticed by macroeconomic policy-makers.

Mutual funds are among active financial institutions in the capital market

and professionally invest the accumulated money of different individuals in

varying markets. As an effective instrument, mutual funds can play a foremost

role in economy management by collecting capitals and leading them to the

industrial sectors of the economy. Mutual funds in developed and developing

countries are continuously growing and increasingly penetrating different

sectors of societies. Statistics show that the movement of the public wealth

towards funds has been recently begun in Iran. Thus, with the ever-increasing

growth of under-management assets, funds roles in the economy, particularly

in the capital market of Iran, have been noticeably valued. Adopting efficient

strategies in cash inflow and outflow management through forming security

portfolios, fund managers have highlighted the status of these institutions in

impacting the prosperity and stability of the capital market indices and other

mutual funds. This spread and development pave the way for the emergence of

systemic risk if it is not accompanied by planning, controlling, and

surveillance.

The joint nature of assets and investment strategies is one of the important

reasons for the investigation of systemic risk in securities mutual funds. Iran

securities mutual funds fall into three categories, including stock, fixed-

income, and balanced mutual funds. The portfolio composition of these

92

Iranian Journal of Finance, 2019, Vol. 3, No. 4 (Shahbazin, F.)

institutions is stocks, fixed-income securities, and bank deposits and each one’s

proportion of investment is varying with concern to the fund type. The

prominent growth of mutual funds and their heightened weights in the Iran

capital market necessitate the systemic risk examination of these financial

intermediaries. In addition, the interconnectivity of mutual funds and their joint

investment nature will cause price changes, derived from the trading behavior

of market activists and the economic conditions of Iran, to influence fund

portfolios. This may bring a shock in the mutual funds, and interconnections

among them, along with an impact on the financial market.

This study specifically pays more attention to the identification of

financial institutions (mutual funds), which more contribute to the emergence

of systemic risk. These institutions are known as systemically important

financial institutions (SIFI). Financial Stability Board (FSB) defines SIFI as

“the financial institutions whose disorderly failure, because of their size,

complexity, and systemic interconnectedness, would cause significant

disruption to the wider financial system and economic activity”. When these

institutions are reckoned as a serious threat to the system, the surveillants and

policy-makers throughout the world ask for the application of more strict

supervisions, including obligations for the excess capital requirement and

liquidity constraints, for such institutions.

This study aims to evaluate and compare the systemic risk and market

risk of mutual funds in Iran. To this end, it first investigates and measures the

systemic risk of Iran mutual funds by the Conditional Value at Risk (CoVaR)

measure and quantile regression approach. Thus, by determining the systemic

risk of each fund, it can identify the mutual funds with higher contribution to

the emergence of systemic risk. The recognition of such financial institutions

can assist surveillants and policy-makers with the exertion of prudential

supervision. After calculating the systemic risk, the funds are categorized into

five classes, and the relationship between systemic risk and market risk is

evaluated by examining diverse variables and employing the analysis of

variance.

In the following and the next section, previous studies conducted on the

modeling and evaluation of systemic risk are reviewed, and the mathematical

model employed in this paper for systemic risk estimation in mutual funds is

described. Then, results are analyzed investigating the relationship between

systemic risk and market risk and discovering systemically important funds.

93


Literature Review

To become more acquainted with the problem of the study, systemic risk, and

measurement methods, some previous studies in this field are outlined.

Adrian and Brunnermeier (2011) were the pioneers who presented a

method known as the Conditional Value at Risk (CoVaR) method for systemic

risk measurement. CoVaR is the same as the Value at Risk (VaR) of a financial

system when financial institutions are under emergency and distressful

conditions. They defined a publicly-traded financial institution's contribution to

systemic risk as the difference between CoVaR conditional on the institution

being under distress and the CoVaR in the median state of the institution

(Adrian & Brunnermeier, CoVaR, 2011). Girardi and Ergün modified the

definition of financial distress in Adrian and Brunnermeier CoVaR to estimate

the systemic risk in the financial industries including depository institutions,

non-depository institutions, insurances, and brokerages, as well as the link

between institutions’ contribution to systemic risk contagion. They found that

the leverage, size, and equity beta are important in explaining the contagion of

systemic risk among financial institutions (Girardi & Ergün, 2013). White, and

et.al. used the vector autoregressive extension to quantile models in order to

estimate systemic risk and analyze spillovers in the Values at Risk between a

market index and financial institutions. The results showed that the long-run

risk of the largest and most leveraged financial institutions is very sensitive to

market-wide shocks in situations of financial distress (White, Kim, &

Manganelli, 2015). Giglio, and et.al. studied the impacts of the 19 varying

measures of the systemic risk and financial market distress on the transference

of shock to the real economy. They used dimension reduction estimators for

constructing systemic risk composite index such that it predicted information

out-of-sample for the lower tail of future macroeconomic shocks (Giglio,

Kelly, & Pruitt, 2016). Kleinow, and et.al. compared the empirical results of

the four conventional and commonly used systemic risk measures, including

marginal expected shortfall, codependence risk, delta conditional value at risk,

and lower tail dependence. Their results indicated that the different criteria of

systemic risk could result in varying risk assessments of various financial

institutions and proposed that the risk assessment of financial institutions

should be carried out accurately and cautiously based on a single risk metric

(Kleinow, Moreira, & Strobl, 2017).

Compared to financial institutions, banks, listed companies in exchanges,

and the real economy, few studies have addressed systemic risk in mutual

funds. Klaus and Rzepkowski employed the Logit model to investigate the

94


spillover effect among hedge funds (Klaus & Rzepkowski, 2009). In a chapter

of the econophysics of systemic risk and network dynamics (Abergel,

Chakrabarti, Chakraborti, & Ghosh, 2012), Abergel, and et.al. adopted a

microscopic network approach to assess the systemic risk of mutual funds.

They found that fund managers could control the systemic risk by prudential

epidemic spread-prevention strategies. Jin and Simone combined marginal

probabilities of distress estimated with the consistent information multivariate

density optimization (CIMDO) methodology and the generalized dynamic

factor model (GDFM) to evaluate systemic risk evaluation in mutual funds (Jin

& Simone, 2014). Pelegrini, and et.al. Calculated the systemic risk of the

money market funds of the UK by the CoVaR approach. The results showed

that liquidity mismatch increased systemic risk in mutual funds (Pellegrini,

Meoli, & Urga, 2017).

Systemic risk measurement does not have a long history in Iran.

However, Sadeghi provided a systemic risk report published by the Research,

Development and Islamic Studies Center of Securities and Exchange

Organization. He theoretically investigated systemic risk in the financial

institutions of the capital market and did not measure systemic risk practically

(Sadeghi, 2012). Ahmadi and Farhanian measured systemic risk by using the

quantile regression approach in 20 listing companies of Tehran Securities

Exchange and considered the interconnectivity among them as a network

(Ahmadi & Farhanian, 2015). Shirmohammadi, and et.al. Investigated systemic

risk among money, insurance, and capital markets and demonstrated a

significant difference between the systemic risk and summation of risk in each

market. Finally, according to the results of the Friedman test, they claimed that

the insurance industry and banking systems had the maximum and minimum

contributions in systemic risk (Shirmohammadi, Chavoshi, & Feshari, 2015).

Moradmand, and et.al. Utilized Delta CoVaR (ΔCoVaR) to measure systemic

risk in 24 firms accepted in the Tehran Securities Exchange. Then, they

employed the Kolmogorov-Smirnov test to rank the stocks of financial firms

based on systemic risk (Moradmand Jalali & Hasanlou, 2016). In their course

of study, Noralidokht and Dadashi-Arani determined the rate of default

contagion in the mass trading network of the capital market. According to their

results, institutions with a maximum impact on the financial network’s

instability had greater relationships with the financial network members or

enjoyed remarkably concentrated associations (Noralidokht & Dadashi Arani,

2015). By selecting the firms that enjoyed the maximum value of the entire

market and outnumbered trading days, Azari-Gharelar and Rastegar measured

systemic risks by using Delta Conditional Value at Risk, expected shortfall;

95


component expected shortfall, and systemic expected shortfall. They showed

that different measures had similar performances (Azari Gharelar & Rastegar,

2015). Mohammadiaghdam, and et.al. Assessed that systemic risk originated

from foreign exchange shocks on financial markets by the Conditional Value at

Risk measure and quantile regression approach (Mohammadiaghdam, Ghavam,

& Fallahshams, 2017). Farzinvash and his colleagues measured systemic risk at

17 banks of the banking network by using the delta CoVaR measure

(Farzinvash, Elahi, Gilanipour, & Mahdavi, 2018). Hekmatifarid, and et.al.

Measured systemic risk in financial markets based on Delta CoVaR

(Hekmatifarid, Rezazadeh, & Malek, 2018).

Rahimi-Baghi, and et.al. Exploited the granger causality network method

to evaluate systemic risk in the financial system of Iran, including banks,

investment firms, and insurances, in the 2011-2017 period. Their results

revealed that the banking and insurance sectors enjoyed the highest and lowest

rates of systemic risk, respectively. Likewise, it became illuminated that the

rate of systemic risk among financial institutions changed over time. The

validation of their research findings demonstrated that the extracted results

were sufficiently valid (Rahimi Baghi, Arabsalehi, & Vaez Barzani, 2019).

Ebadi and his co-authors explored the effect of foreign exchange shock on

the mutual funds systemic risk index by using the multivariate GARCH (M-

GARCH) models and daily net asset value of Iran mutual. The results indicated

that the contagion coefficients of exchange shocks were significant for the

mere return of some funds; however, the presence of contagion among funds

would lead to the spread of the direct effects of exchange shocks through the

transitivity channel of return volatility among funds and increase the funds’

systemic risk index and systemic risk potentiality (Ebadi, Elahi, & Houshmand

Gohar, 2019).

Abrishami, and et.al. analyzed and measured systemic risk in the banking

sector of Iran and examined their influential factors based on three criteria,

including MES, ΔCoVaR, and SRISK, for the listing banks in the stock market.

Their results implied that the Value at Risk of each bank had a positive

relationship to the ΔCoVaR and MES as systemic risk measures. Unlike the

banking literature, not only large banks pose systemic risk, but also small ones

contribute to the genesis and spread of the risk (Abrishami, Mehrara, &

Rahmani, 2019).

Research Methodology

This paper employed the CoVaR measure introduced by Adrian and

96


Brunnemeier (Adrian & Brunnermeier, CoVaR: Dataset, 2016) to assess the

systemic risk of mutual funds in the Iran Capital Market. As the reader is

informed, VaR is the most prevalent risk measure used by financial

institutions, which focuses on the risk of an individual institution. In other

words, is the maximum amount (value) of loss in an event at a

confidence level of q per cent. for an institution is defined as

below:

( ) (1)

Where Xi is the total return of the fund i. The reason for selecting the total

return is that it consists of both fund’s asset price changes and cash dividends

or interest income. The mathematical form of the total return is

( ) ( )

(2)

Where is the total return of mutual fund i at time t, and

( ) are the net asset value of fund i per a unit at times t and t – 1,

respectively, and is the interest income or cash dividends for a unit of fund i

at time t.

The CoVaR is defined as a Value at Risk of the financial system (here a

portfolio of selected funds) conditional on an event ( ) in institution i.

( ) is defined as q-quantile of the conditional probability

distribution.

( | ( ) ( )

) (3)

Where is the total return of fund i, and is the system total return

which is explained as the average of weighted by the net asset value of

funds involved in a given portfolio. More precisely,

∑

(4)

In which is the weight of fund i among N funds that are present in a

system or portfolio.

97


There are two different definitions of Conditional Value at Risk (CoVaR)

in various references based on their condition ( ( )). The symbol

is the main initial definition introduced by (Adrian & Brunnermeier, CoVaR,

2011), and displays q-quantile of the system return ( ) conditional

on , while

is a newer CoVaR definition introduced by

Girardi and Ergun (2013), in which the conditional term is .

Generally, and

are defined as the q-quantile of the

following conditional distributions:

(

) (5)

(

) (6)

It should be noted that the values and q, as the confidence levels of the

fund and system can vary; however, in this paper, we consider both of them

equal to 5 per cent. Since this study used the method of Adrian and

Brunnermeier (2011), calculations were carried out based on Eq. (5).

(Adrian & Brunnermeier, CoVaR, 2011) Measured the contribution of

each financial institution to the systemic risk by - i.e., the difference

between - provided that an institution had a distress condition, and

in the median or normal conditions of the institution.

( ) ( )

( )

(7)

Quantile Regression

As observed in Eq. (5), the calculation of CoVaR requires the estimation of the

VaR for each fund and the portfolio of funds. Despite various methods that

have been proposed for VaR estimation, quantile regression was employed in

this paper.

Let ( ) be a vector of state variables. Then,

quantile regression is defined as

(8)

Where the residual term satisfies in ( | ) . ( ) is the

98


conditional quantile function.

By estimating the coefficient vector ( ) of Eq. (8), the VaR can be

estimated for fund i (or system) with a confidence level of as

(9)

In the next step, CoVaR is calculated by quantile regression and rewritten

similar to Eq. (8) as

(10)

Using the estimated value of ,

can be estimated.

Particularly, can be estimated by employing quantile regression as

(11)

Hence,

is calculated as

(12)

Where

is the maximum loss of a system with confidence level

q per cent when the fund i encounters a loss of a per cent, and

reveals fund declination when institute i moves from the median state to the

worst scenario of per cent. For the time estimation of VaR, these measures

will be defined in the forms of time-varying measures.

Results

The present study employed the library and documentary methods for data

collection. For this purpose, Persian and English electronic resources were

exploited. The requisite data for the study variables were collected from the

Rasam System, Part Financial Data Processing Co., and Tehran Securities

Exchange. Finally, statistical techniques were applied for the examination of

the research hypotheses.

The data included the weekly data of all mutual funds and exchange-

traded funds (i.e., stock, fixed-income, and balanced funds) licensed by the Iran

Securities and Exchange Organization (SEO) that they were active from early

March 2017 to late September 2019 and enjoyed the information in 75 per cent

99


of the period. The data consisted of the weekly net asset values and total

returns of funds. Moreover, the data obtained from the Tehran Stock Exchange

total return index (TEDPIX) and Iran Farabourse yield to maturity of the

Islamic treasury bills were considered as state variables.

The total number of mutual funds was 193 throughout this period. With

respect to the conducted refinement, 41 funds were excluded, and 152

remained. Table 1 shows the number of mutual funds separated by the fund

type.

Table 1. Number of Understudy Mutual Funds Separated by Fund Type

Type of fund Fixed-income Stock Balanced Total

before filtration 79 93 21 193

after filtration 70 64 18 152

Table 2 provides the descriptive statistics of the variables employed in the

study. This study used the weighted return of the net asset value as the system

return ( ), while the variations in TEDPIX and yield to maturity of the

treasury bills were considered as the state variables.

Table 2. Descriptive Statistics of Variables

Mean on

funds' total

return

Mean on

Weighted

return

TEDPI

X

return

YTM

variatio

n

Mean of NAV

(Billion IRR)

average 0.64 0.29 1.15 -0.07 9,443

standard

error 0.12 0.01 0.27 0.28 47

standard

deviation 1.38 0.11 3.08 3.2 548

median 0.53 0.27 0.65 -0.19 9,408

minimum -3.03 -0.01 -9.61 -9.34 8,319

maximum 7.05 0.87 13.81 15.48 10,771

range 10.08 0.89 23.42 24.82 2,451

number of

observation 131 131 131 131 131

The simple mean of the weekly returns of 152 understudy funds was 0.64

per cent during 131 weeks ending up with late September 2019 (i.e., a period

of two years and six months). However, the net asset value-weighted average

return of the funds was 0.29 per cent. In this period, the weekly return mean of

TEDPIX was 1.15 per cent, and its minimum and maximum rates were -9.61

100


per cent and 13.81 per cent, respectively. The yield to the maturity of treasury

bills had an average variation of -0.07 per cent, and its variation range was

24.82 per cent during the period. The net asset value mean of the funds was

9,443 billion IRR.

Figure 1 illustrates the graphs of the variations in the weekly return of the

fund's system, the weekly return of TEDPIX, the weekly yield to the maturity

of treasury bills, and the net asset value mean of the mutual funds refined

during 131 weeks ending up with late September 2019.

(a)

(b)

(c)

(d)

Figure 1. The Graphs of a) Mean Variations in Weekly Return of System, b) Weekly Return

TEDPIX, c) Weekly Variation Yield to Maturity of Treasury Bills, and d) Net Asset Value

Mean of Mutual Funds

The results derived from the quantile regression approach shows that

Asemen Yekom and Armaghan Iranian mutual funds were systemically

important in late September 2019. During the year ending up with the same

101


date, Bourseiran and Gohar Nafis mutual fund were systemically important.

Tables 3 and 4 provide the list of top five systemically important mutual funds

in late September 2019, the year ending up with the same date.

Table 3. Five Systemically Important Funds in the Late September 2019

Rank Fund Name Fund Type

1 Aseman Yekom stock

2 Armaghan Iranian fixed-income

3 Bazr Omid Afarin stock

4 Yekom Iranian fixed-income

5 Etemad Meli Bank fixed-income

As can be seen in Table 3, stock funds Aseman Yekom and Bazr Omid

Afarin and fixed-income funds Armaghan Iranian, Yekom Iranian, and Etemad

Meli Bank were identified as systemically important funds in late September

2019.

Table 4. Five Systemically Important Funds in the Year Leading to Late September 2019

Rank Fund Name Fund Type

1 Bourseiran stock

2 Gohar Nafis Tamadon balanced

3 Ofogh Mellat stock

4 Dey Bank stock

5 Tajrobe Iranian balanced

Bourseiran, Gohar Nafis Tamadon, and Ofogh Mellat mutual funds were

three important funds in the year ending up with late September 2019. In other

words, the system experienced the maximum CoVaR when the funds were

distressed, i.e. the fund return was less than its VaR.

Tables 5 and 6 report the quantile regression results for the Value at Risk

of Aseman Yekom mutual fund and the Conditional Value at Risk of the same

fund in distress conditions in late September 2019, respectively.

It should be noted that in every phase (131 weeks) and for every fund

(152 funds), four regressions are fitted for the Conditional Value at Risk

(CoVaR).

102


Table 5. Quantile Regression Results for Value at Risk of Aseman Yekom Mutual Fund in

Late September 2019

coefficients t-stats standard error p-value

Constant -8.84 -4.26 2.07 0.00

TRI volatility 0.36 0.78 0.46 0.44

YTM volatility 1.78 1.46 1.22 0.15

Table 6. Quantile Regression Results for Conditional Value at Risk of System in Late

September 2019 When Aseman Yekom Mutual Fund is in stress

coefficients t-stats standard error p-value

constant -0.02 -0.4 0.06 0.69

Fund's return 0.01 0.51 0.01 0.61

TRI volatility 0 0.16 0.02 0.87

YTM volatility 0.05 1.8 0.03 0.08

Figure 2 demonstrates the over-time trend of the quantile regression

coefficients for CoVaR calculation when the Aseman Yekom mutual fund is

distressed.

Figure 2. Over-Time Trend of Quantile Regression Coefficient

As can be seen, the changes in the constant-coefficient were sensible over

time, changing from positive values at the beginning of the period to negative

ones in late September 2019. The changing behavior of the total return index

(TRI) and fund return (Xi) coefficients is noticeable; an increase in one

coefficient reduces another.

To develop a better prospect from quantile regression estimation in VaR

and CoVaR calculations, Figure 3 illustrates the weekly return of Burseiran

mutual fund, its VaR, and the systemic effect of the fund on the entire system.

103


(a)

(b)

Figure 3. The Graphs of a) Weekly Return and VaR of Bourseiran Mutual Fund, b) Weekly

System return, CoVaR, and ΔCoVaR from Early March 2017 to Late September 2019

This paper has so far evaluated the systemic risk of mutual funds in the

Iran Capital Market and estimated the VaRs of the funds as a measure of

market risk. To compare the systemic risk and market risk of mutual funds, the

next phase assigns the funds to five rank-ordered categories’; the funds falling

into the first group are the most systemically important, while the funds in the

fifth group are the least systemically important. This categorization was carried

out weekly, and some variables, including total return, net asset value, VaR,

beta coefficient of systematic risk, CoVaR, and ΔCoVaR, were investigated in

each fund group.

The purpose of this categorization was to examine the relationships

between the systemic risk of mutual funds and each one of the above-

mentioned variables. Since the funds were rank-ordered based on ΔCoVaR, the

mutual funds in the first-class apparently enjoyed higher CoVaR and ΔCoVaR

values than the others. Figure 4 depicts the graphs of the variables in each

class.

104


(a)

(b)

Figure 4. The Graphs of a) Mean CoVaR in the Five Groups, and b) Mean ΔCoVaR in the Five

Groups from Early March 2017 to Late September 2019

Figure 5 demonstrates the mean number of mutual funds separated by the

fund type in each class. As can be seen, the number of mutual funds, including

stock and fixed-income funds, has a reverse trend, such that the stock funds in

classes with higher systemic risk outnumber the fixed-income funds, and vice

versa. The number of mutual funds in every class is nearly 30.

Figure 5. Mean Number of Mutual Funds Separated by Fund Type in Each Class

Concerning the above results, the behavior of the parameters present in

each class is expected to be dependent on the behavior of their included funds.

For example, the beta coefficient and VaR of the high-class funds are expected

to be larger than those of low-class funds. Figure 6 shows the beta coefficient

and VaR scatter plots for classes 1, 3, and 5 compared to CoVaR.

105


(a)

(b)

Figure 6. The Scatter Plots of a) Beta Coefficient and ΔCoVaR and b) VaR and ΔCoVaR for

Classes 1, 3, and 5

Figure 6a shows the scatter plot of the beta coefficient relative to

. In class 1, which is depicted in blue, the ΔCoVaR shows more

negative values, and the beta coefficient, implying systematic risk, is larger. In

class 5, shown in yellow, the beta coefficients are smaller than other classes,

and the approach zero.

A similar behavior is observed in Figure 6b, which depicts the VaR

scatter plot versus with .

Table 7 summarizes the mean parameters within the classes to obtain

more accurate analyses concerning the behavior of the classes.

Table 7. Mean Parameters of Funds in the Classes

Class 1 Class 2 Class 3 Class 4 Class 5

Net Asset Value Mean (Billion IRR) 5752 7091 10193 10681 15132

Delta CoVaR -0.28 -0.24 -0.20 -0.16 -0.10

VaR -3.48 -2.81 -1.93 -1.31 -1.04

Beta 0.64 0.59 0.44 0.30 0.23

The net asset value mean has an uptrend form in Class 1 compared to

Class 5 due to the presence of many fixed-income funds in less systemically

important classes. As can be seen, there is intuitively a significant relationship

between the classes with varying systemic risk and other individual risk

parameters such as VaR and beta; the funds with riskier VaR and beta

106


coefficient fall into a more systemically important fund class.

For the more meticulous investigation of the relationships between the

parameters in different classes, the one-way analysis of variance (ANOVA)

was employed to discover whether there were significant relationships between

the mean values of more than three independent groups. The null hypothesis of

this test is

Where is the group’s mean, and is the number of groups, which was 5

in this study. If the one-way analysis of variance offers statistically significant

results, the alternative hypothesis will be accepted – i.e., there are minimally

two groups with significantly different means.

It is worth noting that the one-way analysis of variance is a one-way test

and cannot discern which group’s mean value differs. To this end, it is required

to test both groups simultaneously. The MATLAB machine learning toolbox

was employed for this purpose.

Figure 7 shows the graph of beta coefficient variations in the classes. As

can be seen, the beta mean has a downtrend in the classes.

Figure 7. The Bar Chart of Beta Coefficient Mean in Five Classes

The results of the one-way analysis on the beta coefficient parameter in

five classes show the presence of minimally two different classes with distinct

means. Table 8 outlines the test results.

107


Table 8. One-Way Analysis of Variance for Beta Coefficient Variable

Source of

variation

Sum of

Squares (SS)

Degree of

Freedom (df)

Mean

Squares (MS)

F-

Stats

Prob

>F

columns 6.812827 4 1.703207 38.1

05

4E-

25

error 11.62155 260 0.044698

sum 18.43438 264

The F-statistic and its p-value imply the presence of minimally two

classes with distinct means. Table 9 summarizes the analysis of variance test

for the two distinct classes. As can be seen, the means of Classes 1 and 2 and

those of Classes 4 and 5 were not different at a confidence level of 95%, while

the means of other classes were significantly different, which confirms the

claim concerning the presence of significant relationships between the beta

coefficients of the groups. This indicates that there is a significant relationship

between systemic risk and the beta coefficient as a market risk variable of

funds.

Table 9. One-Way ANOVA of Beta Coefficient Mean in Pairwise Classes

First

Group

Second

Group

95 % confidence interval for Mean Mean

Difference p-Value

Lower bound Upper bound

1 2 -0.054 0.058 0.170 0.620

1 3 0.096 0.208 0.320 0.000

1 4 0.234 0.346 0.458 0.000

1 5 0.305 0.417 0.529 0.000

2 3 0.038 0.150 0.262 0.002

2 4 0.176 0.288 0.400 0.000

2 5 0.247 0.359 0.471 0.000

3 4 0.026 0.138 0.250 0.007

3 5 0.097 0.209 0.321 0.000

4 5 -0.041 0.071 0.183 0.421

108


Figure 8 displays the boxplot of beta coefficient variations in the five classes.

Figure 8. Boxplot of Beta Coefficient Variations in Five Classes

The central line of the boxplot (red line) represents the median, the box’s

edges indicate the 25%and 75% quantiles, and the separated lines display

extreme values, which are not reckoned as outlier data. The remaining outlier

points were drawn separately at the line ends. In Figure 8, the deviation of data

in every class can be observed.

Figure 9 displays the graph of changes in VaR for the classes. As can be

seen, the VaR mean has an uptrend in each class.

Figure 9. The Bar Chart of Value at Risk Mean in Five Different Classes

The results of the one-way analysis of variance of the VaR in the five

classes imply the presence of minimally two different classes with distinct

means. Table 10 outlines the results of this test.

109


Table 10. One-Way Analysis of Variance of Value at Risk

Source of

variation

Sum of

Squares (SS)

Degree of

Freedom (df)

Mean

Squares (MS)

F-

Stats

Prob

>F

columns 220.87310 4 55.21828 43.52

778

6.08E

-28

error 329.82963 260 1.268575

sum 550.70273 264

The F-statistic and its p-value indicate the presence of minimally two

classes with distinct means. Table 11 summarizes the analysis of variance test

for each pair of classes. As can be seen, the means of Classes 4 and 5 do not

differ at a confidence level of 95%, while the means of other classes are

significantly different.

Table 11. One-Way Analysis of Variance of Value at Risk in Pairwise Classes

First

Group

Second

Group

95 % confidence interval for Mean Mean

Difference p-Value

Lower bound Upper bound

1 2 -1.267 -0.671 -0.074 0.019

1 3 -2.149 -1.552 -0.956 0.000

1 4 -2.760 -2.164 -1.567 0.000

1 5 -3.034 -2.437 -1.840 0.000

2 3 -1.479 -0.882 -0.285 0.001

2 4 -2.090 -1.493 -0.896 0.000

2 5 -2.363 -1.767 -1.170 0.000

3 4 -1.208 -0.611 -0.014 0.042

3 5 -1.481 -0.885 -0.288 0.001

4 5 -0.870 -0.274 0.323 0.722

110


Figure 10 displays the boxplot of Value at Risk in the five classes.

Figure 10. Boxplot of Value at Risk in Five Different Classes.

Conclusion and Suggestions

The quantile regression approach is a conventional market-value based method

to calculate Conditional Value at risk (CoVaR) as a systemic risk measure. By

making use of this method, the present study endeavored to measure the

systemic risk of mutual funds industry, and by categorizing them into different

classes, it investigated the behavior of different funds in every class and

compared its relationship with the market risk. The numerical results indicated

that stock mutual funds were systemically more important than the other funds,

including fixed-income and balanced funds, due to their high return volatility,

which makes them riskier. The categorization and analysis of variance test

determined that the beta coefficient mean and Value at Risk (VaR), as the

indicators of funds’ market risk, had direct relationships with the funds’

systemic risk; a higher the systemic risk of a fund stood for a higher beta

coefficient and VaR. In other words, the results of variance analysis verified

the significant relationship between the systemic risk and market risk of a fund

(beta systematic risk and VaR). This result approves of the result of Abrishami,

and et.al work (Abrishami, Mehrara, & Rahmani, 2019) that found a positive

(direct) relationship between VaR and Delta CoVaR. Concerning the results, it

is suggested that researchers evaluate the systemic risk of funds by other

conventional measures and compare the results to discover systemically

important mutual funds.

111


References

Abergel, F., Chakrabarti, B., Chakraborti, A., & Ghosh, A. (2012).

Econophysics of systemic risk and network dynamics. Springer Science & Business

Media.

Abrishami, H., Mehrara, M., & Rahmani, M. (2019). Measuring and Analysis of

Systemic Risk in Iranian Banking Sector and Investigating Its Determinants.

Econometric Modeling, 11-36.

Adrian, T., & Brunnermeier, M. (2016). CoVaR: Dataset. American Economic

Review.

Adrian, T., & Brunnermeier, M. K. (2011). CoVaR. . CoVaR (No. w17454).

National Bureau of Economic Research.

Ahmadi, Z., & Farhanian, M. (2015). Systemic Risk Measuring in Tehran Stock

Exchange with CoVaR and MES Approaches. Journal of Securities Exchange, 7(26),

3.

Azari Gharelar, A., & Rastegar, M. (2015). Master thesis on A Comparision

Among Systemic Risk Measures in Tehran Stock Exchange. Tehran, Iran: Kharazmi

University.

Ebadi, J., Elahi, N., & Houshmand Gohar, S. (2019). Effect of Exchange Rate

Change Shocks on Systemic Risk Index Among Mutual Funds. Economic Research

and Policies, 27(89), 373-398.

Farzinvash, A., Elahi, N., Gilanipour, J., & Mahdavi, G. (2018). The evaluation

of Systemic Risk in the Iran Banking System by Delta Conditional Value at Risk

(CoVaR) Criterion. Financial Engineering and Portfolio Management, 8(33), 265-281.

Giglio, S., Kelly, B., & Pruitt, S. (2016). Systemic risk and the macroeconomy:

An empirical evaluation. Journal of Financial Economics, 119(3), 457-471.

Girardi, G., & Ergün, T. (2013). Systemic risk measurement: Multivariate

GARCH estimation of CoVaR. Journal of Banking & Finance 37.8, 3169-3180.

Hamid Abrishami, M. M. (2019). Measuring and Analysis of Systemic Risk in

Iranian Banking Sector and Investigating Its Determinants. Econometric Modeling,

11-36.

Hekmatifarid, S., Rezazadeh, A., & Malek, A. (2018). The Estimation of

Systematic Risk in Iranian Financial Sectors (ΔCoVaR Approach). Quarterly Journal

of Economic Modelling, 12(3), 99-122.

Jin, X., & Simone, F. (2014). Banking systemic vulnerabilities: A tail-risk

dynamic CIMDO approach. Journal of Financial Stability(14), 81-101.

112


Klaus, B., & Rzepkowski, B. (2009). Risk spillover among hedge funds: The

role of redemptions and fund failures. European Central Bank.

Kleinow, J., Moreira, F., & Strobl, S. (2017). Measuring systemic risk: A

comparison of alternative market-based approaches. Finance Research Letters 21, 40-

46.

Mohammadiaghdam, S., Ghavam, M., & Fallahshams, M. (2017). Assessment

of the Systemic Risk Originated from the Currency Shocks in the Financial Markets of

Iran. Financial Research, 19(47), 475-504.

Moradmand Jalali, M., & Hasanlou, K. (2016). The Assessment of Share of

Banks, Insurance and Investment Companies in Systemic Risk. Quarterly Journal of

Islamic Finance and Banking Studies, 67-92.

Noralidokht, S., & Dadashi Arani, H. (2015). Master thesis on Contagion

Robustness in Financial Networks. Zanjan, Iran: Institute for Advanced Studies in

Basic Sciences (IASBS).

Pellegrini, C., Meoli, M., & Urga, G. (2017). Money market funds, shadow

banking and systemic risk in the United Kingdom. Finance Research Letters 21, 163-

171.

Rahimi Baghi, A., Arabsalehi, M., & Vaez Barzani, M. (2019). Assessing the

Systemic Risk in the Financial System of Iran using Granger Causality Network

Method. Financial Research, 21(53), 121-142.

Sadeghi, M. (2012). Systemic risk management in Iran financial institutions.

Research, Development and Islamic Studies Department of Iran Securities and

Exchanges Organization.

Shirmohammadi, F., Chavoshi, S., & Feshari, M. (2015). Master thesis on

systemic risk in Iran financial market. Tehran: Iran University of Economic Science.

White, H., Kim, T.-H., & Manganelli, S. (2015). VAR for VaR: Measuring tail

dependence using multivariate regression quantiles. Journal of Econometrics 187.1,

169-188.

Bibliographic information of this paper for citing:

Shahbazin, Fereshteh; Ghalibaf Asl, Hasan; Seyghali, Mohsen & Peymani Foroushani,

Moslem (2019). Evaluating and Comparing Systemic Risk and Market Risk of Mutual

Funds in Iran Capital Market. Iranian Journal of Finance, 3(4), 90-112.

Copyright © 2019, Fereshteh Shahbazin, Hasan Ghalibaf Asl, Mohsen Seyghali and

Moslem Peymani Foroushani

Documents

Evaluating and Comparing Systemic Risk and Market Risk of